Criteo Privacy Preserving ML Competition
The Online Advertising industry is seeing a major shift today in its operational constraints with a global movement towards more privacy. Popular techniques for privacy-compliant advertising such as aggregation and differential privacy mechanisms were shown to match high privacy standards but also raise concerns about the possibility to learn relevant machine learning models for ad placement.
We propose in this challenge to explore the trade-off between privacy level and prediction performance, on data donated by Criteo - an industry leader that already released several open datasets for research purposes. To anchor the competition in reality, the challenge design is inspired by (and as close as possible/convenient to) current propositions in the Privacy Sandbox discussed in the Improving Web Advertising forum at W3C.
The competition consists in two tasks of click and sales prediction under privacy constraints, for a total prize money of $20,000.
The competition website is accessible at http://go.criteo.net/criteo-privacy-preserving-ml-competition.